GraphDB Lucene Connector - uppercase problem when indexing URIs - lucene

Question

There seems to be a strange behaviour with GraphDB Lucene connectors when a URI is indexed (either because of using the $self quasiproperty or because the property chain leads to a URI). I would summarize the issue as follows:
Uppercase letters in the URI must be escaped in the query text (i.e. query text must be "*\Merlo" instead of "*Merlo" in the wine example provided here)
No snippet can be extracted from URIs
Any idea how this could be overcome?

Answer 1

The Lucene connectors treats URIs as non-analyzed fields, i.e. as a single chunk of string whose tokens aren't words and aren't tokenized or analyzed as words. The logic is that URIs are identifiers and they have a meaning only in their entirety. URIs don't contain text, even if they sometimes make sense to people. This also means that normal full-text queries will not work on such fields, nor any snippets can be extracted. They can be search but since it's not full-text they might behave unexpectedly.
In your particular example with a query like "*Merlo", Lucene will run the query through the analyzer in order to be able to match analyzed fields (i.e. the fields normally used for full-text search). By escaping the capital letter you're preventing the analyzer from normalizing it to a lowercase m and you get a match.
If you need an exact match you can do this too (note there's no need to escape the capital letters):
PREFIX : <http://www.ontotext.com/connectors/lucene#>
PREFIX inst: <http://www.ontotext.com/connectors/lucene/instance#>
SELECT * {
?search a inst:my_index ;
# Surround with double quotes to force an exact match
:query "\"http://www.ontotext.com/example/wine#Merlo\"" ;
:entities ?entity .
}